“Holy schmoly! What are we going to do about this darn graph?”
Don’t panic! Just follow these simple instructions…
Choose your cutoffs
“The trend shows no increase for the last [n] days/months/years.”
Don’t mention the previous [20 n] data points.
Talk about the trend of the trend
“Sure the graph is going up for now, but the rate of increase is going down.”
If this fails, talk about the rate of increase of the rate of increase. Keep on differentiating until you find a curve that matches your needs. If all else fails, try logarithms.
Count the phases
“There are just as many downward phases as upward phases.”
There will always be as many up phases as down phases (plus or minus one). Make sure you distract listeners from the magnitude of the phases.
Choose an outlier
“If there really is an upward trend, how can we explain the result from Convenienceville?”
In any large body of noisy data there will be subsets that do not fit the overall trend. Be sure to remove all error bars from your graph as these will give the game away.
“There is so much doubt about the trend that we can’t say anything until we have more evidence.”
Evidence is your friend. More evidence means more cutoffs to choose from, more trends to analyse, more phases to count, more outliers to discover, and more confusion to sow. Be careful to disguise the fact that you and not the data are the source of the confusion.